CLIM Fall 2008 What are the Roles of Satellites & Supercomputers in Studying Weather and Climate? CLIM 101
CLIM Fall 2008 Satellites and Supercomputers Satellites –Observing the weather –Observing the climate and planetary ecosystems –Provide global coverage –Provide nearly continuous observations Supercomputers –Modeling the atmosphere –Numerical weather prediction –Modeling the physical climate system –Modeling the Earth system –Simulating the current and future climate
CLIM Fall 2008
Geostationary Satellite SMS-A, 1974: 1st geostationary
CLIM Fall 2008 GOES-8 - Geostationary Earth Orbiting Satellite
CLIM Fall 2008 GOES East - Full Disk View
CLIM Fall 2008 GOES West - Zoomed View
CLIM Fall 2008 Mosaic of Multiple Geostationary Satellite Images (infrared band)
CLIM Fall 2008 Geostationary Satellite Polar Orbiting Satellite SMS-A, 1974: 1st geostationary TIROS-1, 1960: 1st polar orbiter
CLIM Fall 2008
How are weather forecasts made? by Wiley Miller
CLIM Fall 2008 Numerical Weather Prediction 1.Determine (continuous) equations to be solved –Equation of state or Ideal Gas Law (Boyle’s Law relates P V, Charles’ Law relates V T, Gay-Lussac’s Law relates T P) –Conservation of mass (dry air, water) –Conservation of energy –Conservation of angular momentum –Result: set of coupled, nonlinear, partial differential equations 2.Discretize the equations for numerical solution (typically requires computer) 3.Measure current state of global atmosphere to obtain initial conditions 4.Solve the initial value problem to produce a forecast 5.Take into account uncertainty in measured atmospheric state by repeating step 4 over an ensemble of slightly different initial conditions
CLIM Fall 2008 (approximation) Mass conservation Energy conservation Newton’s law = p / p s
CLIM Fall 2008 Using the secant approach, we can approximate the temperature equation: By reducing t, we can obtain an increasingly accurate solution.
CLIM Fall 2008 Discretization Atmosphere and ocean are continuous fluids … but computers can only represent discrete objects
CLIM Fall 2008 Discretization Atmosphere and ocean are continuous fluids … but computers can only represent discrete objects
CLIM Fall 2008 John von Neumann Seymour Cray & Cray-1 ENIAC IBM 360 Cray-2 Columbia NASA
CLIM Fall 2008 Computing Capability & Global Model Grid Size (km) Peak Rate:10 TFLOPS100 TFLOPS1 PFLOPS10 PFLOPS100 PFLOPS Cores (1 st year available) 1,400 (2006) 12,000 (2008) ,000 (2009) ,000 (2011) 6,000,000? (20xx?) Global NWP 0 : 5-10 days/hr (~20X10 6 points) Seasonal 1 : days/day (~20X10 6 points) Decadal 1 : 5-10 yrs/day (~2X10 6 points) Climate Change 2 : yrs/day (~0.5X10 6 points) Range: Assumed efficiency of 10-40% 0 - Atmospheric General Circulation Model (AGCM; 100 levels) 1 - Coupled Ocean-Atmosphere-Land Model (CGCM; ~ 2X AGCM computation with 100-level OGCM) 2 - Earth System Model (ESM; ~ 2X CGCM computation) * Core counts above O(10 4 ) are unprecedented for weather or climate codes, so the last 3 columns require getting 3 orders of magnitude in scalable parallelization (scalar processors assumed; vector processors would have lower processor counts) Thanks to Jim Abeles (IBM)
CLIM Fall 2008 Numerical Weather Prediction: Data Assimilation Cycle
CLIM Fall 2008
NCEP Forecast Ensembles: 850 hPa Geopotential Height Contours: 126 and 147 gpm Valid 03Oct08 Target1-day forecast3-day forecast4-day forecast2-day forecast
CLIM Fall 2008
Mobile Radar Hurricane Katrina Intensity at Landfall 29 Aug Z 4 km WRF, 62 h forecast Courtesy of P. Fox (NCAR)
CLIM Fall 2008